Accuracy assessment of airborne laser scanner dataset by means of parametric and non-parametric statistical methods

Parametric and non-parametric statistical methods are compared and discussed for the accuracy assessment of digital surface models coming from airborne laser scanner. Such datasets are chosen from different types of areas: from the relatively favourable flat terrain to the complex built-up areas and abrupt terrain zones with vegetation. For those complex sites, a reference `ground truth´ is established using a mobile mapping system, whereas flat terrain areas are checked through static ground surveying using global positioning system. The experimental results confirm that systematic errors and many outliers remain in the airborne laser scanner dataset corresponding to built-up and abrupt terrain areas, and thus that nonparametric statistical methods are suitable for an accuracy assessment.